SOURCENESTSource · Sourcing & spend intelligence

Know whether a quote is fair before you negotiate, and stop paying more than you agreed.

SourceNest is sourcing and spend intelligence for procurement teams. It builds a fair-price band from your own purchase history and public price drivers, finds money leaking through off-contract buying and duplicate invoices, and proposes award splits that a category head approves.

Run the demo Discuss your sourcing challenge

Stacked steel coils in a covered storage yard

SourceNest at a glance

1Reads

Purchase orders, goods receipts and invoices from your ERP
Rate contracts with their price tables, clauses and end dates
Public price drivers such as WPI paper and board, RBI exchange rates and energy prices

2Decides

SourceNest

Is this quote below, within or above the fair band?
Which leaks to investigate and recover
How to split the award across suppliers

3Produces

Award recorded in SourceNest once a category head approves
Negotiation round opened in the supplier portal
Recommendation package prepared for your ERP; the buyer raises the PO

↺ Learns Savings ledger: identified, accepted, implemented, realized back to 1 · Reads

The business problem

A 9% price rise, and nobody can say if it is fair

Your corrugated-box supplier writes on a Friday. Kraft paper is up, so boxes go up 9% from next month. The line is worth about ₹4 crore a year. The buyer has last year's price, a phone and a week.
Nobody can say what a box should cost today. Paper indices moved, but by how much for this grade? Meanwhile Plant 2 has been invoiced above the contract rate since June, one invoice was paid twice, and a plant buyer has been ordering from a local converter outside the contract.
The rise is negotiated down to 6%. Everyone calls it a saving. Nobody checks it against what the invoices show three months later.

Why the usual fixes fall short

Spreadsheets hold last year's price. They cannot say what this year's price should be after paper, resin or metal indices moved.
Three-way match confirms that the invoice agrees with the PO and the goods receipt. If the PO itself was raised above the contract rate, or with no contract at all, the match still passes.
Sourcing suites run the RFQ, but the award is still split by judgement, and the saving is booked the day it is agreed rather than when invoices prove it.

Supplier increases keep landing at whatever the negotiation achieves, and money already paid above contract stays paid.

Gross marginInput-cost rises pass through fast and cuts pass through slowly. Every point on packaging and raw material comes straight off margin.
CashDuplicate payments and invoices above the contract rate leave the business and are rarely recovered.
Supply riskChasing the lowest bid can put a whole line on one fragile supplier.
Trust with financeProcurement reports negotiated savings. Finance sees no change in cost. The two numbers never meet.

In plain terms

What SourceNest is, and how it fits

What it is: A sourcing and spend-intelligence layer beside your ERP. It reads purchase history, contracts and invoices, and gives buyers a fair-price view, a list of leaks and an award proposal.
Who uses it: CPOs and heads of procurement, category managers and buyers, with the CFO as co-sponsor for the savings ledger. Suppliers quote through a secure link.
What it does: Classifies spend, checks quotes against a should-cost band, finds five kinds of spend leakage, runs sealed-bid RFQs and e-auctions, proposes the lowest-cost award split within your supply rules, prepares the negotiation, and tracks each saving from identified to realized.
How it fits your systems: It reads from SAP, Tally with GST-IRP, or Operix, read-only first. Connectors are configured per deployment. It never raises a purchase order. Your buyer does that in your ERP after approval.
What is different: The numbers come from deterministic maths and an optimization solver, never from a language model. Every recommendation carries its evidence and abstains when data is thin. Savings are reported in stages, so a modeled number is never shown as banked.

See it work

SourceNest in action

A box supplier asks for 9% more. Follow the quote from the fair-price check to an approved award and the invoices that test it.

A simulated scenario on representative data. You choose the option and give the approval; the numbers follow your choices.

End-to-end workflow

From data to decision to outcome

What goes in, what SourceNest works out, what gets decided, what happens next, and how the result comes back.

InputsDATA

Purchase orders, goods receipts and invoices from your ERP
Rate contracts with their price tables, clauses and end dates
Public price drivers such as WPI paper and board, RBI exchange rates and energy prices
Sealed supplier quotes and documents through the supplier portal
Your supply rules: MOQ, maximum share, lead-time cap, number of suppliers

IntelligencePREDICT

Should-cost regression on public drivers, giving an 80% fair-price band
Rules-first spend classification and five leakage detectors, each with ₹ impact and cited invoice lines
Techno-commercial scoring of sealed bids, placed against the fair-price band
Award optimizer on Google OR-Tools CP-SAT
Supplier 360 scorecard with a continuity circuit-breaker

DecisionsDECIDE

Is this quote below, within or above the fair band?
Which leaks to investigate and recover
How to split the award across suppliers
Negotiation target, fallback and walk-away price
When to renew a contract, and whether to buy now or defer

ExecutionEXECUTE

Award recorded in SourceNest once a category head approves
Negotiation round opened in the supplier portal
Recommendation package prepared for your ERP; the buyer raises the PO
Leakage findings opened for investigation, with the evidence accounts payable needs for recovery

FeedbackLEARN

Savings ledger: identified, accepted, implemented, realized
Quoted price compared with the price on later invoices
Every human override stored in Procurement Memory
Fair-price band checked against prices actually paid

Modules

Capabilities, organized by the work they do

14 modules in 4 groups. Open a module for its inputs, outputs, users and an example.

Price intelligence3 modules Spend and leakage4 modules Sourcing and award5 modules Proof and memory2 modules

Price intelligence

Know what a fair price is for each material, and walk into the negotiation with a plan.

ONI fair-price band and quote checkTells a buyer whether a quote is below, within or above a fair-price band for that material today, and which cost drivers explain the price.

Challenge: Buyers negotiate from last year's price because nobody can say what the price should be after input costs moved.
Inputs: Your own PO, GRN and invoice prices (at least six months for the should-cost band), plus free public driver series such as WPI paper and board, RBI exchange rates and energy prices.
Outputs: Two views side by side: your own price index from what you actually paid, and an 80% should-cost band from public drivers. A below, within or above verdict with the % gap to the midpoint, a breakdown of the price by driver, and the evidence behind it. It abstains when history is thin or the band fails its calibration check.
Used by: Buyers, category managers
Value: Negotiations start from a number the buyer can defend, labeled as your cost intelligence rather than a market price.
Works with: Driver series are refreshed per deployment from free public sources. Packaging goes to grade level first: paper, board and corrugation.
In action · illustrative: A box quote of ₹46.2/kg against a band of ₹41.0–44.2/kg reads "above band, 8.5% over the midpoint". The breakdown shows kraft paper as the driver that moved, and by how much less than the quote implies.

Buy-timing signalSuggests buy now, defer or neutral for a material, and only speaks when it beats a simple baseline.

Challenge: Forward buys made on a hunch tie up cash or miss a cheaper month.
Inputs: The material's price history and its driver series.
Outputs: A defer, buy-now or neutral signal, tested against a naive "price stays the same" baseline.
Used by: Category managers, finance
Value: Forward buys are timed on evidence, and the signal stays neutral when it has no edge.
Works with: —
In action · illustrative: Kraft paper drivers have eased for three months. The signal says defer the quarterly forward buy, and shows how it scored against the baseline.

Negotiation packGives the buyer a target, a fallback, a walk-away price, the best alternative if talks fail and a short call script before the negotiation.

Challenge: Without a cost-based target, the supplier's first counter-offer becomes the deal, and preparing one properly takes a senior buyer hours.
Inputs: The fair-price band, the supplier's price history, sealed bids from other suppliers and continuity limits from Supplier 360.
Outputs: Target, fallback and walk-away prices in ₹ per unit with the reasoning for each, the best alternative to a deal, and talking points. It abstains when the fair-price band abstains.
Used by: Buyers, category managers
Value: Every negotiation starts from the same evidence, whoever runs it.
Works with: The prices are computed by deterministic maths. A language model writes the script and the reasoning only, never the numbers.
In action · illustrative: For the box line: target ₹43.6/kg, fallback ₹44.2/kg, walk away above that. Shree's sealed bid at ₹43.2/kg is the alternative, and the script opens on what kraft paper actually did.

Spend and leakage

See where the money goes, and find what is already leaking with the evidence to recover it.

Spend analytics and classificationClassifies every purchase line into your categories and shows spend by category, supplier, plant and month, including the long tail.

Challenge: Spend sits in PO lines with free-text descriptions and inconsistent codes, so nobody can say what the business buys, from whom, and at what price.
Inputs: PO, GRN and invoice lines from your ERP, material and supplier masters, and your category tree.
Outputs: A spend cube by category, supplier, plant and month; ABC tail-spend analysis; and a category for each line, with uncertain lines marked for review.
Used by: CPO, category managers, procurement analysts
Value: Every other check starts from a complete, correctly classified picture of spend.
Works with: Rules classify first. A language model suggests a category only for lines the rules cannot place, and a person can correct any line.
In action · illustrative: "5PLY RSC 450x300 BRN" and "Corrugated shipper 5-ply" land in the same category, so box spend reads as one ₹4.2 crore line instead of four smaller ones spread across plants.

Leakage detectorsFinds five kinds of leak in purchase data: off-contract buying, maverick buying, prices above contract or benchmark, duplicate invoices and PO prices above the contract rate.

Challenge: These leaks pass three-way match because the PO, receipt and invoice agree with each other, even when the price breaks the contract. Most are found at year-end, if ever.
Inputs: PO and invoice lines, rate contracts with their price tables, and your preferred supplier per category.
Outputs: One standard alert per supplier and material with the ₹ impact, the cited invoice lines, the contract it breaks, an evidence score and a recovery action. Nothing is recovered automatically.
Used by: Buyers, category managers, accounts payable
Value: Leaks surface with the evidence accounts payable needs to recover them. Thresholds are set per company: by default a 2% price overage and a 7-day duplicate window.
Works with: Scans run on a schedule over spend read from your ERP. When later invoices confirm a recovery, the savings ledger moves it forward without writing anything back to the ERP.
In action · illustrative: Plant 2 was billed ₹43.9/kg on a ₹42.4/kg contract for 38 tonnes. One alert, ₹57,000, with the invoice lines cited.

Supplier and material data checkFinds duplicate vendors and inconsistent material and contract records before they distort the analysis.

Challenge: The same supplier exists three times under different names, so its spend and leaks are split and missed.
Inputs: Supplier master with GSTIN, PAN and names; material master; contracts.
Outputs: Duplicate-vendor findings matched on GSTIN, PAN and name similarity, a completeness score and contract-versus-master mismatches. It never merges records by itself.
Used by: Procurement operations, master-data team
Value: Spend and leakage views count each supplier once.
Works with: Runs on the same schedule as the leakage scans.
In action · illustrative: "Shree Packaging Pvt Ltd" and "Shri Packaging" share a PAN. The pair is flagged for the master-data team to merge in the ERP.

Contract intelligencePulls clauses and obligations out of contracts, warns before renewals and expiries, and flags invoices billed above the contracted rate.

Challenge: Contracts lapse or auto-renew at stale rates, and invoices quietly bill above the agreed price.
Inputs: Contracts with their rate tables, dates and documents; PO and invoice lines.
Outputs: Clauses and obligations, each cited to the passage in the source document; renewal and expiry alerts; and contract-versus-PO over-bill findings passed to the leakage register.
Used by: Category managers, legal, accounts payable
Value: Renewals start early, while there is still time to run a proper RFQ, and over-billing is caught with the clause in hand.
Works with: Clause extraction reads the document with a language model, cites the source passage for every clause and abstains when the text is unclear. Renewal alerts and over-bill checks are rules.
In action · illustrative: The box rate contract ends in six weeks. A renewal alert is raised with the spend under it and its price-adjustment clause, and the over-bill check shows Plant 2 billed above the rate that clause allows.

Sourcing and award

Run the RFQ, compare sealed bids and split the award within your supply rules, with a person signing.

RFQ, sealed bids and e-auctionRuns RFQs, RFPs and e-auctions with sealed bids, and compares them side by side on price, lead time, MOQ, capability and fairness against the ONI band.

Challenge: Quotes arrive by email in different formats, early ones can leak to later bidders, and comparison is done by hand.
Inputs: RFQ lines, invited suppliers, scoring weights and the fair-price band for each material.
Outputs: Sealed submissions and, once the bid closes, a ranked techno-commercial comparison with each bid placed against the fair-price band.
Used by: Buyers, category managers
Value: Every supplier quotes on the same terms, and the comparison can be repeated by anyone.
Works with: —
In action · illustrative: Three box makers quote through their own links. After sealing, the comparison scores them 92, 84 and 61, and shows which bids sit inside the fair band.

Supplier portalGives each supplier a secure link to quote, upload documents and answer negotiation rounds.

Challenge: Supplier documents and counter-offers live in inboxes, so nobody can show who said what, and when.
Inputs: Invites, quotes, and certificates such as COA and MSDS.
Outputs: Quote submissions, a document exchange, negotiation rounds and a tamper-evident log in which each entry is hash-chained to the one before.
Used by: Buyers and external suppliers
Value: A complete record of the negotiation that audit and the supplier can both rely on.
Works with: —
In action · illustrative: The incumbent receives a round-two request; its reply and timestamp land in the hash-chained log.

Award optimizer and approval gateProposes the lowest-cost split of an award across suppliers within your supply rules, and moves it from proposed to approved to handed off.

Challenge: Awards are split by habit, or all volume goes to the lowest bid and supply becomes fragile.
Inputs: Sealed bids, demand to cover, MOQs, minimum and maximum share, lead-time cap and maximum number of suppliers.
Outputs: A proposed split with its modeled ₹ saving, computed in exact decimal arithmetic, and a plain-language rationale. An award is never executed on its own. It abstains when a rule cannot be modeled.
Used by: Buyers, sourcing approver
Value: Award splits are cost-optimal, explainable and inside the rules the business set.
Works with: Runs on Google OR-Tools CP-SAT. A sourcing approver signs every award; above the ₹50 lakh modeled-saving ceiling it escalates and nothing is written.
In action · illustrative: 1,000 tonnes of boxes, a 60% share cap and a 21-day lead-time limit: 60% to a proven alternate, 40% to the incumbent. The category head signs it before anything is handed off.

Supplier 360 and circuit-breakerOne view of each supplier with delivery, quality, price-drift and compliance scorecards, a continuity-of-supply score, and a circuit-breaker that stops awards which would leave you on one fragile supplier.

Challenge: Single-source risk and slipping OTIF stay invisible until a line stops, and savings from consolidation vanish the first time a sole supplier misses a delivery.
Inputs: Deliveries (OTIF), quality results, price history, share of spend, GST and MSME status, and certificates such as COA.
Outputs: Scorecards, a risk score, a continuity score with single-source exposure in ₹, and a circuit-breaker state of clear, warning or tripped, with a dual-source recommendation when it trips.
Used by: Category managers, quality
Value: Cost and continuity are weighed together before an award is approved.
Works with: Feeds the award optimizer and the negotiation pack.
In action · illustrative: A sole supplier holding 78% of a category with a weak continuity score trips the breaker. The next award must add a second source, even if the optimizer found a cheaper single-source split.

Sourcing copilotDrafts RFQs, explains an award in plain language and answers supplier questions, with every statement cited to its source.

Challenge: RFQ drafting and award write-ups take senior buyers hours, and supplier questions pile up in email.
Inputs: Past RFQs and specifications, the award and the optimizer's output, ONI bands, supplier history and documents.
Outputs: A cited first draft of the RFQ, an "explain this award" note for approvers and finance, and answers to supplier questions from the RFQ documents. Any action it proposes goes through the approval gate.
Used by: Buyers, category managers, approvers
Value: Senior buyers spend their time on judgement, and every write-up carries its evidence.
Works with: A language model writes the prose through the AI gateway. Numbers come from ONI and the optimizer, never from the model.
In action · illustrative: Asked to explain the box award, it writes that Shree takes 60% because its sealed bid is the lowest within the 21-day lead-time rule, and the incumbent keeps 40% for continuity. Each sentence links to the bid, rule or scorecard behind it.

Proof and memory

Track every saving to the invoice, and keep the reasoning behind every decision.

Savings realization ledgerTracks each saving through four stages: identified, accepted, implemented and realized.

Challenge: Procurement reports negotiated savings that finance never sees in cost.
Inputs: Approved recommendations, later POs and invoices, and signed contracts.
Outputs: A ledger by stage with a modeled-versus-realized flag, a CFO view of savings since go-live traced to each approved action, and flags on savings that have gone stale.
Used by: CFO, CPO, finance analysts
Value: One savings number that procurement and finance agree on.
Works with: A saving moves to realized only on a downstream signal: a PO or invoice at the new price, or a signed contract. Nothing is written back to your ERP.
In action · illustrative: The box award enters as identified at ₹18 lakh a year, finance accepts it, and it counts as realized only once invoices at the new price arrive.

Procurement MemoryRecords every suggestion, human override and outcome, so similar decisions later show what was done before and how it turned out.

Challenge: A category manager's judgement leaves with them, and the same debate repeats every year.
Inputs: Suggestions, overrides, dismissals, approved decisions and their measured outcomes.
Outputs: A memory event for each edit, with who made it and what changed; similar past cases at the next decision; and recommendations re-ranked by the outcomes they produced.
Used by: Buyers, category managers
Value: Judgement builds up over time instead of resetting with each new buyer.
Works with: Each company's memory is its own and is never shared. Re-ranking uses transparent weights and abstains until there is enough history.
In action · illustrative: The approver moved a split from 60/40 to 70/30 for lead-time comfort. Next year's box award shows that note and what happened after.

Intelligence and automation

What the intelligence actually does

Five levels, from seeing what is happening to learning from it.

Visibility

Spend by category, supplier and plant, with every leak listed, its ₹ impact and the invoice lines behind it, and a 360 view of each supplier.

Prediction

A fair-price band for each material from your price history and public drivers, a buy-now, defer or neutral timing signal, and renewal warnings before contracts lapse.

Decision

The lowest-cost award split under your MOQ, share, lead-time and supplier-count rules, with a negotiation target, fallback and walk-away price.

Execution

After approval the award is recorded, the supplier round is opened and a recommendation package is prepared for your ERP. SourceNest never raises a purchase order; your buyer does.

Learning

Each saving is tracked toward realized against later invoices, and every override is stored so the next similar decision shows what was done before.

Runs automatically

Fair-price band and quote verdict, including the abstain when history is thin
Spend classification and leakage scans over spend read from your ERP
Bid scoring and award-split calculation
Duplicate-vendor detection (it never merges records by itself)
Renewal and expiry alerts

Always waits for a named person

Every award, by a sourcing approver such as the category head; above the ₹50 lakh modeled-saving ceiling it escalates and nothing is written
Every purchase order, raised by your buyer in your ERP
Recovery on a leakage finding: investigate or dismiss is a person's call
Accepting a saving into the ledger, by finance

Business value

What it helps you work toward, and how we measure it

We agree the baseline with you in writing before a pilot starts. These are the outcomes we measure.

Fewer increases accepted above fair

Buyers open each negotiation with a band and a target instead of last year's price.

Share of accepted price changes inside the fair band, by category, against a baseline agreed in writing.

Cash recovered from leakage

Overpaid and duplicate amounts are found with the evidence needed to claim them back.

Amounts confirmed and recovered, by detector, from the leakage register.

More spend on contract

Off-contract and maverick buying is visible plant by plant.

Off-contract and maverick spend as a share of category spend, month on month.

Savings finance agrees with

A saving is reported by stage, so modeled and banked numbers are never mixed.

Savings by ledger stage (identified, accepted, implemented, realized), reconciled with finance each quarter.

No line on one fragile supplier

Awards weigh cost and continuity together.

Lines above your agreed supplier share, and circuit-breaker warnings resolved.

Illustrative example

Illustrative: a ₹4.2 crore box line, a quote of ₹46.2/kg and a fair band of ₹41.0–44.2/kg. Settling at the top of the band instead of the quote is worth about ₹20 lakh a year on 1,000 tonnes. The ledger reports it as identified until invoices at the new price arrive.

Enterprise readiness

Integration, control and deployment

ERP & system integration

Connectors for SAP (MM and FI tables over OData), Tally with GST-IRP and Operix, read-only first. Write-back is limited to a recommendation package and, if you switch it on, draft requisitions. Connectors are configured per deployment.

Data sources & connectivity

Uses your own PO, GRN, invoice and contract history plus public price drivers, refreshed per deployment. Cross-company benchmarks need consent and at least five companies in a cohort.

Deployment & implementation

Six services with Postgres and object storage for supplier documents, run as a managed service. The hosting region is agreed per deployment.

Configuration & business rules

Approval ceiling, leakage thresholds, circuit-breaker limits and award rules are set per company. Paper & board, apparel, FMCG, pharma, paints and agri packs are switched on per company.

Roles, access & approvals

Roles for CPO, category manager, buyer, finance viewer and admin, with a separate scope for supplier users. Sign-in uses your identity provider.

Security, governance & audit

Tenant ID on every table with row-level security policies. The supplier portal log is SHA-256 hash-chained, with update and delete revoked. No security certification is claimed.

Human oversight & exceptions

A sourcing approver signs every award; above the ₹50 lakh ceiling it escalates and nothing is written. Recommendations carry an evidence score and basis, and can abstain. Language models write prose only.

Scale & ongoing optimization

Built for many companies on one platform, separated by tenant.

Straight answers

Questions about SourceNest

Something else? Ask a specialist.

Does SourceNest place purchase orders?

No. It recommends, and records the award once a person approves it. Your buyer raises the PO in your ERP. It is built so that it cannot create a PO by itself.

Where does the fair price come from?

From your own purchase history (POs, goods receipts and invoices) plus public price drivers such as WPI paper and board and RBI exchange rates. It is an 80% band, presented as your cost intelligence and not as a market benchmark. With fewer than six months of history, or if the band fails its calibration check, it abstains.

Does an AI decide the award?

No. The split comes from an optimization solver (Google OR-Tools CP-SAT) working within your MOQ, share, lead-time and supplier-count rules, and a category head approves it. Language models only write drafts and explanations. They never produce the numbers.

Which systems does it work with?

SAP, Tally with GST-IRP, and Operix. It reads first, read-only, and connectors are configured per deployment. The only things it writes back are a recommendation package and, if you switch it on, draft requisitions, always after approval.

When does a saving count?

A saving moves through four stages: identified, accepted, implemented and realized. Realized is reserved for savings confirmed by a downstream signal, such as invoices or a PO at the new price, or a signed contract.

Next step

Discuss your sourcing challenge

SourceNest goes after the money procurement tools tend to miss: quotes above fair, invoices above contract, payments made twice. Every number carries its evidence, and every award carries a named approver.

Talk to a specialist

Manufacturers and FMCG companies with heavy material and packaging spend, running SAP, Tally or Operix, who want savings proven in rupees. Industry packs cover paper & board, apparel, FMCG, pharma, paints and agri inputs.

In the same loop